Fujitsu Unit Proposed Monetizing Idle Network Compute
The 1Finity platform aims to let operators rent out underutilized RAN capacity to enterprises for AI tasks.
Updated on Oct. 5, 2026 in Artificial Intelligence
Live Poll
Do you trust telecommunications companies to provide reliable enterprise AI services alongside their network connectivity?
Fujitsu company 1Finity has proposed a model for network operators to sell idle compute capacity within Radio Access Networks (RAN) to enterprise customers. The effort aims to improve infrastructure ROI by utilizing the significant underutilized headroom inherent in networks designed for peak capacity.
Why it matters
Operators face challenges justifying infrastructure investments based strictly on network performance improvements, leading them to seek new revenue streams from existing assets. By leveraging their physical advantages in last-mile connectivity and edge distribution, they aim to compete with hyperscalers for enterprise AI workloads.
A demonstration showcased shoplifting detection using a chain of CCTV, CPE, and an Open RAN-compliant radio powered by 3 NVIDIA GPUs. The system utilizes Armada software to monitor RAN-specific KPIs and enforce policy-driven orchestration, ensuring network traffic maintains priority over AI workloads.
The players
1Finity
A subsidiary of Fujitsu that focuses on network infrastructure solutions and AI-RAN integration.
Aible
A software provider that develops enterprise-ready AI application agents for automated business tasks.
Armada
A resource management software provider specializing in tracking and optimizing network-specific key performance indicators.
NVIDIA
A designer of graphics processing units that serve as the primary compute architecture for modern AI workloads.
Fujitsu
A multinational technology corporation with a legacy in hardware, software, and global telecommunications infrastructure.
The details
The system relies on policy-driven orchestration, a management layer that dynamically governs resource allocation, to prevent enterprise AI tasks from degrading network performance. By using self-service models, operators allow enterprises to deploy over 100 Aible application agents—pre-configured software modules for business logic—directly on the network edge. This architecture leverages existing RAN infrastructure, which is designed to handle maximum traffic spikes and therefore maintains significant unused processing capacity during normal operations.
Timeline
October 5, 2026: The initiative was presented at the Intelligent RAN Forum.
The Tech Race
The effort marks a strategic shift to transform radio infrastructure into an AI-native service model, moving beyond basic connectivity. It pits network operators against hyperscalers by highlighting the operator's inherent advantage in local edge site distribution.
For enterprise users, this model promises access to compute power closer to the edge, potentially lowering latency for AI applications like computer vision. Actual availability and pricing are currently unannounced as the technology moves from demonstration to operator adoption.
The takeaway
The move suggests a future where radio networks function as decentralized data centers rather than just data pipes. Stakeholders should watch for operator pilots that integrate these AI-RAN services into standard enterprise offerings.
Further reading
Explore the latest shifts in Artificial Intelligence to see how network-integrated compute compares to existing cloud models.
Live Poll
Do you trust telecommunications companies to provide reliable enterprise AI services alongside their network connectivity?
